Reconstructing patchy reionization with deep learning

The precision anticipated from next-generation cosmic microwave background (CMB) surveys will create opportunities for characteristically new insights into cosmology. Secondary anisotropies of the CMB have an increased importance in forthcoming surveys, due both to cosmological information they encode and role play obscuring our view primary fluctuations. Quadratic estimators become standard tools reconstructing fields that distort produce secondary anisotropies. While successful lensing reconstruction with current data, quadratic be suboptimal other effects at expected sensitivity upcoming surveys. In this paper we describe a convolutional neural network, ResUNet-CMB, is capable simultaneous two sources anisotropies, gravitational patchy reionization. We show ResUNet-CMB network significantly outperforms estimator low noise levels not subject lensing-induced bias on reionization would present straightforward application estimator.